Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
A Tale of Two Tropes:
On the Relationship Between Metaphor and Simile
Catrinel Haught
Rider University
Running Head: DIFFERENCES BETWEEN METAPHORS AND SIMILES
Address: Rider University
Department of Psychology
Science Hall, 320A
2083 Lawrenceville Road
Lawrenceville, NJ 08648
Email: [email protected]
Phone: (609) 895-5689
Acknowledgements: I thank Sam Glucksberg and Ray Gibbs for their helpful comments.
DIFFERENCES BETWEEN METAPHORS AND SIMILES
2
2
Abstract
Four experiments tested three competing theories of metaphor comprehension:
comparison, categorization, and career-of-metaphor. The findings shed light on key
mechanisms involved in metaphor processing and conceptual combination. They show
that some novel tropes are privileged in metaphor over simile form, and others may
express different interpretations in simile and in metaphor form. These results speak
against the assumption that metaphors and similes are interchangeable, thus providing
support for the categorization model. A unifying account of metaphor comprehension is
proposed, along with a discussion of implications for conceptual combination.
New metaphors are quickly created and propagated among the speakers of any language
and even across languages. The experiments described below addressed the issue of how
novel and conventional metaphors and their corresponding similes are understood and
interpreted.
Existing models propose different mechanisms for metaphor comprehension (see,
for example, Gibbs, 2011; Gibbs & Colston, 2012; Steen, 2007). Comparison theorists
(Ortony, 1979; Gentner, 1983; Fogelin, 1988) claim that metaphors such as “Some
lawyers are sharks” are understood as implicit similes, e.g., “Some lawyers are like
sharks”, via a property matching process. Proponents of the categorization model
(Glucksberg & Keysar, 1990; Glucksberg, 2001; Glucksberg & Haught, 2006; Haught,
2013) argue that metaphors are understood as categorization statements via dual-
DIFFERENCES BETWEEN METAPHORS AND SIMILES
3
3
reference, an interactive property attribution process. Similes, on the other hand, are
understood as comparisons. In metaphor form, the vehicle term shark refers to an
abstract level, to the class of vicious, predatory creatures, of which the literal shark is a
prototype. In simile form, it is used to refer to the literal level, to the marine creature
with fins and leathery skin. The career of metaphor account (Bowdle & Gentner, 2005)
aims to integrate the two approaches. It proposes that novel metaphors are always
processed as comparisons, i.e., similes. As these metaphors become conventionalized
through repeated use, they can be processed as comparisons or categorizations.
A more detailed overview of each of these models, below, is followed by a
description of four experiments that tested the predictions of these models.
The Comparison Model
The proponents of standard pragmatic theory assume that nominal metaphors (of
the type X is a Y) must first be recognized as false literal assertions, which violate the
maxim of truthfulness in a conversation (Grice, 1975). Indeed, if taken literally, the
statement that some lawyers are sharks is “defective” (Searle, 1979), i.e., it does not
make sense. Once this violation has been detected, an active search begins for an
alternative, non-literal, interpretation that is truthful and informative. According to this
model, which applies not only to metaphors, but to any type of non-literal language,
people always attempt to first derive the literal meaning of an utterance. When the literal
meaning of a metaphor such as “Some lawyers are sharks” does not satisfy the maxims
of conversation, the statement is converted into a simile: “Some lawyers are like sharks”,
which can then be interpreted as any other literal comparison. Thus, a literally false,
DIFFERENCES BETWEEN METAPHORS AND SIMILES
4
4
“defective” statement becomes a comparison, i.e., a statement that is always true (because
one can always identify ways, whether obscure or obvious, in which two things are
alike).
The comparison view of metaphor comprehension postulates that metaphors are
recognized as implicit similes and understood via a comparison process (Gentner, 1983;
Gentner & Wolff, 1997; Miller, 1979; Ortony, 1979). Thus, nominal metaphors such as
“Some lawyers are sharks” are treated as if they were explicit comparisons, i.e., “Some
lawyers are like sharks”. In this nominal metaphor of the form X is a Y, some lawyers
(the X term) is the topic of the metaphor, and sharks (the Y term) is the vehicle. When
treated as a simile, the features of the topic are compared with the features of the vehicle
and only the relevant subset of these features is used in interpreting the given statement.
How are these common features identified?
Ortony (1979) has proposed a salience imbalance mechanism, based on Tversky’s
(1977) contrast model of similarity, to explain feature selection in metaphor
comprehension. According to Tversky’s theory, perceived similarity between two
objects X and Y is a weighted function of common features that are shared by X and Y
minus the features that distinguish X from Y, i.e., are not common to both terms. In an
elaboration on Tversky’s contrast model, Ortony put forth his account of metaphor
recognition and comprehension based on the notions of a) knowledge representations,
whose function is to hold the information regarding the attributes of a concept, and b)
salience imbalance. His proposal rests on the assumption that the same feature can have
different levels of salience depending on the particular concept of which it is an attribute.
DIFFERENCES BETWEEN METAPHORS AND SIMILES
5
5
The referents of a comparison X is like a Y can share features that differ in terms
of their salience, yielding four possible kinds of matches: 1) The matching properties can
be salient for the predicate term Y, but not for the subject term X. 2) They can be of low
salience for both terms, e.g., brown and creamy for “Chocolate syrup is like steak sauce”.
3) They can be of high salience for both terms, as in “Billboards are like placards”. Or
4) they can be highly salient for the subject term X, but not salient for the predicate term.
Assertions whose shared properties are described in the first two cases do not convey any
new information, hence they are perceived as anomalous. Statements that fit under the
third description, i.e., whose features are highly salient for both the X and the Y terms,
are understood as literal comparisons. For example, in the statement “Billboards are like
placards”, both terms have equally salient matching properties. The fourth case
describes the requirements for a statement to be understood as a figurative comparison.
For example, in the statement “Some billboards are like warts”, shared features such as
unsightly are highly salient for the predicate warts, but are not salient for the subject
concept billboards. In this last case, the salience constraint is also the criterion used to
judge the level of metaphoricity of an utterance: "the imbalance […] in salience levels of
matching attributes of the two terms is a principle source of metaphoricity" (Ortony,
1979, p. 164).
Ortony proposed salience imbalance as the criterion for distinguishing figurative
statements from literal ones and specified the importance of informativeness in accepting
and interpreting a statement. However, literal assertions of the type described in case 3
above, i.e., high salience for A and B, as in “Billboards are like placards”, do not meet
the informativeness requirement: they convey information that is trivial to anyone other
DIFFERENCES BETWEEN METAPHORS AND SIMILES
6
6
than, perhaps, a listener who does not share the same high salience match as the speaker
(a situation which is difficult to imagine). Indeed, for any comparative statement to be
considered informative, the matching properties must be of low salience for the subject
term X and highly salient for the predicate term Y. Ortony reserves this salience
imbalance scenario to metaphoric statements, but upon closer observation, following the
informativeness rule, it appears to be a requirement for any comparison statement,
whether literal or metaphoric. Therefore, the model fails to account for distinctions
between the literal and the figurative.
Despite its intuitive appeal, Ortony’s simple matching mechanism is not an
adequate explanatory framework for the comprehension of either literal statements or
metaphors, the latter of which are treated as implicit comparisons, i.e., similes. The
model fails to account for those utterances, literal or metaphoric, that convey new
information about the subject. For instance, the metaphoric statement “Peter is like a
shark” informs the listener about a set of properties that characterize Peter, but that were
not part of the listener’s knowledge about this person before the metaphor was processed
(Glucksberg, 2000).
Gentner (1988) also argued that metaphors are understood as comparisons. She
proposed an elaboration of Ortony’s answer to the question of how the features of
metaphor interpretation are identified. According to her structure-mapping model, the
matching features are often relations, not object attributes, as the salience imbalance
model had suggested. An example used by Gentner is Rutherford’s analogy between the
hydrogen atom and the solar system: in this metaphor, the physical attributes of the
entities are irrelevant. Instead, the relationship between the planets and the sun is
DIFFERENCES BETWEEN METAPHORS AND SIMILES
7
7
mapped onto the domain of the topic. This account was initially developed as an
explanatory framework for a limited subset of analogies, but was then extended as an
account for metaphor comprehension. Indeed, the application of the structure-mapping
model to metaphor comprehension assumes that metaphor is a type of analogy.
According to the initial account, a metaphor’s topic and vehicle can be matched in three
ways: in terms of their relational structure, e.g., “The moon is a lightbulb”; in terms of
their attributional structure, e.g., “A cloud is a marshmallow”; or in terms of both
relational structure and attributes, e.g., “Tree trunks are drinking straws” (the relation of
upward movement of liquid and the attribute of thin, tube-like appearance).
According to the revised structure-mapping theory, in metaphors, relations that
apply to the vehicle also apply to the topic and these relations are independent of whether
the objects represented by the topic and vehicle are similar or not (Gentner et al., 2001;
Gentner & Markman, 1997). In other words, similarity is based on relations. Consistent
with this notion, when participants are asked to generate interpretations of metaphors,
their paraphrases include more relations than properties (e.g., Gentner & Clement, 1988;
Shen, 1992; Tourangeau & Rips, 1991).
According to both Ortony’s and Gentner’s accounts, the grounds for comparison
are established via a matching process. Properties (or relations) of the metaphor topic
and vehicle are first extracted exhaustively and then checked one against the other. From
the set of matching properties (or relations), those that are most relevant and informative
are identified and used as the grounds for the comparison. The ensuing interpretation
process is then no different than the one used for literal comparisons.
DIFFERENCES BETWEEN METAPHORS AND SIMILES
8
8
The Categorization Model
Glucksberg and Keysar (1990) proposed that nominal metaphors are processed
not as comparisons, but as class inclusion statements, with the metaphor vehicle serving
as a name for the superordinate category to which the topic is assigned. In the example
“Some lawyers are sharks”, the metaphor vehicle (sharks) is used to refer to a
superordinate category (‘predatory and vicious creatures’) within which the metaphor
topic (some lawyers) is included. The vehicle term sharks has dual reference, i.e., it can
be used at the literal level to denote the marine fish shark or at the metaphorical level to
denote the category of ‘predatory, ferocious creatures’ that is exemplified by the literal
term sharks. In a similar vein, Utsumi and Sakamoto (2011) proposed categorization as
the key mechanism in understanding predicative metaphors, and Chiappe, Kennedy and
Smykowski (2003) showed that metaphors are distinct from similes, and function more
like categorization statements.
Dual reference is not unique to figurative language, and is a common
communicative strategy. In literal language, words such as Kleenex or Frisbees, which
represent brand names, are commonly used to refer to superordinate categories, e.g., all
tissue products and all flying discs. The dual reference property is especially useful in
classifier languages, such as the American Sign Language, that do not have names for
superordinate categories, but that do have basic level names which can serve as
prototypes of a category (Glucksberg, 2001). In such languages, the nameless
superordinate category can still be referred to through a word used to represent the
overall category. Even in English, not all categories have lexicalized names of their own,
in which case a noun referring to a prototypical member of the class can be used to refer
DIFFERENCES BETWEEN METAPHORS AND SIMILES
9
9
to the superordinate category as a whole (as in the Kleenex example above). In
Glucksberg and Keysar’s (1990) example “My job is a jail”, the vehicle jail is considered
to be a prototypical member of the class of restrictive and unpleasant situations, and it
can therefore be successfully employed to name the superordinate category of such
confining places or circumstances to which the topic job belongs.
Dual reference explains why similes such as “Some ideas are like diamonds” can
be paraphrased as metaphors, e.g., “Some ideas are diamonds”, but literal comparisons
such as “Emeralds are like rubies” cannot be paraphrased as category assertions, e.g.,
“Emeralds are sapphires”. In literal comparisons, the two concepts are at the same level
of abstraction and belong to the same category, e.g., both emeralds and rubies are gems.
Indeed, if a new concept X were introduced by comparing it to an existing concept, e.g.,
X is like a sapphire, one would infer that it is part of the same superordinate category,
e.g., gems, to which the predicate sapphire belongs. In literal comparisons, the predicate
term does not have any other level of abstraction beyond its literal referent, so
paraphrasing such a comparison as a categorization statement is not permissible. In
metaphoric assertions, the vehicle refers to both the literal and the abstract referent,
which allows for paraphrases of metaphors as similes and of similes as metaphors.
However, the fact that the referents of the simile and metaphor predicates are distinct also
allows for the possibility of differences in interpretation of the two tropes, i.e., metaphors
and similes may not be always be identical paraphrases of each other.
According to the categorization model, metaphor comprehension involves
property attribution (Glucksberg & Keysar, 1990; 1993; Glucksberg, McGlone, &
Manfredi, 1997) rather than property matching, as comparison theorists had proposed.
DIFFERENCES BETWEEN METAPHORS AND SIMILES
10
10
Two important criteria are used to determine which vehicle properties are attributed to
the metaphor topic (Glucksberg, 2000). One is a superordinate category to which the
vehicle term refers and which is specified by dual reference. The other criterion is the
relevance of that category’s prototypical properties to the topic, which is a function of the
interaction between vehicle attributes and constraining topic dimensions.
This second representational assumption assigns different roles to the vehicle and
topic terms, with the vehicle yielding properties for attribution to the topic and the topic
yielding the relevant dimensions for attribution. In the lawyers/sharks metaphor, the
topic lawyers provides different relevant dimensions such as competence, reputation and
cost. The vehicle sharks provides candidate properties such as vicious, predatory, that
can be attributed to the topic. In this example, the vehicle term sharks has a limited
number of highly salient properties that characterize the category of predatory, aggressive
creatures. The topic lawyers also has relatively few relevant dimensions that highly
constraint the attribution process. Clearly, not all metaphors have vehicles with such few
salient vehicle attributes and topics with such few, highly restricting relevant dimensions.
Metaphors topics such as my parents have many relevant dimensions, hence they pose
minimal constraints on the attribution process. Metaphors can also have vehicles that
differ as a function of the number or type of candidate properties. In some instances, the
same literal vehicle exemplar, e.g., snake, can be prototypical of more than one category,
e.g., “anything with twisting shape” or “anything devious.” The dual reference, property
attribution framework proposed by the categorization model can account for all these
instances, regardless of the number of vehicle attributes and the level of constraint
imposed by the topic. For example, in the metaphoric context “Some lawyers are
DIFFERENCES BETWEEN METAPHORS AND SIMILES
11
11
snakes”, the topic lawyers constrains the candidate attributes to those that are meaningful
for topic dimensions such as character. In the metaphor “Some roads are snakes”, shape,
rather than character, is a relevant dimension for the topic roads, hence the attribution is
restricted to properties of snakes such as twisting, curved form.
The Career of Metaphor
Bowdle and Gentner (Bowdle & Gentner, 2005; Gentner & Bowdle, 2001) have
put forth a hybrid model, called the “career of metaphor.“ It claims that the processing of
metaphors shifts from comparison to categorization, as metaphors are used repeatedly
and thus become conventionalized. According to this account, conventionalization is a
property of the vehicle term, not the entire metaphorical expression. In the lawyer-shark
example, shark would be a conventional vehicle for “devious, aggressive creature.”
Novel metaphors, such as “Science is a glacier” or “Mind is a kitchen”, have
vehicles, e.g., glacier and kitchen, with a clear literal sense, but without a well established
metaphoric sense. Therefore, these assertions can only be processed as comparisons. In
conventional metaphors, e.g., “An opportunity is a doorway”, vehicles have been used
repeatedly and therefore now have both a literal and an abstract, metaphoric meaning.
This polysemy or dual reference allows for the possibility that they can be interpreted as
either comparisons or as categorizations.
Although the dual reference function proposed by Glucksberg and his colleagues
is incorporated into their hybrid account, Bowdle and Gentner argue that metaphor
comprehension does not depend on categorization, but on the type of alignment process
described in the structure-mapping model. All novel metaphors are understood as
DIFFERENCES BETWEEN METAPHORS AND SIMILES
12
12
comparisons and only conventional metaphors can (although do not have to) be processed
via categorization. When conventional metaphors are processed as categorizations, the
metaphoric categories to which the vehicles refer take the form of abstract relational
schemas and they are nothing but a byproduct of figurative comparisons.
The career of metaphor argument was supported by data from studies that
compared novel tropes, such as “Mind is (like) a kitchen” and “A newspaper is (like) a
telescope” to conventional tropes, such as “Faith is (like) an anchor” and “Alcohol is
(like) a crutch”. For instance, in a ratings task, participants strongly preferred the simile
form for novel assertions, but showed no preference for similes or metaphors for
conventional assertions. A similar effect was observed in a study that measured speed of
processing: novel tropes were understood more quickly when presented as similes than as
metaphors, but the reverse was true for conventionalized tropes (Bowdle & Gentner,
2005).
The hypothesized shift from comparison to categorization was assessed most
directly in an experiment that attempted to simulate the process of metaphor
conventionalization (Bowdle & Gentner, 2005). Participants were exposed to sets of
novel similes with the same vehicle, e.g., “An acrobat is like a butterfly” and “A figure
skater is like a butterfly.” Then they were asked to generate a topic for the statement
“_______ is like a butterfly” such that it expresses the same meaning as in the previous
two similes. The test phase measured participants’ expressed preference for the metaphor
form, e.g., “A ballerina is a butterfly”, relative to the simile form, e.g., “A ballerina is
like a butterfly”. Metaphor preference was greater when participants had been exposed
to the vehicle concept than when they had not. Assuming that the repeated exposure to
DIFFERENCES BETWEEN METAPHORS AND SIMILES
13
13
the same simile predicate simulated the conventionalization of metaphors, the results
were taken as support for a shift from comparison to categorization. While this finding
does indicate an increase in categorization as the mode of processing for conventional
metaphors, it does not show that novel metaphors are understood only via comparison.
Another problematic issue is the confounding role of aptness in the experiment
that compared comprehension times for conventional and novel metaphors and similes.
Conventional tropes were understood more quickly in metaphor than in simile form, but
the reverse was true for novel tropes, which took longer to understand in metaphor than
in simile form. This finding was attributed to differences in conventionality and was
interpreted as support for the claim that novel tropes could be processes only as
comparisons, whereas conventional ones could be processed as categorizations.
However, the items used in this experiment also appeared to differ in aptness, with novel
metaphors such as “Mind is a kitchen” or “The fisherman was a spider”, seeming less apt
than the conventionalized ones, e.g., “An opportunity is a doorway” or “Alcohol is a
crutch”. It is therefore difficult to assess how much of the variability reported in the
results is attributed to the confounding variable of aptness and how much is the results of
the intended manipulation, i.e., a metaphor’s level of conventionalization. Bowdle and
Gentner acknowledged this issue and they conducted a post-hoc test to measure the
aptness of the experimental items employed. The novel tropes were indeed rated
significantly less apt than the conventional ones. Within the novel tropes, similes were
rated as significantly more apt than the corresponding metaphors, a finding that Bowdle
and Gentner interpret as support for their account that all novel metaphors are initially
understood as comparisons, i.e., similes. However, this result does not eliminate the
DIFFERENCES BETWEEN METAPHORS AND SIMILES
14
14
confounding effect of aptness. In fact, it may well reinforce it: because the novel
figuratives are less apt, they are more likely to be processed as comparisons than as
categorizations.
Interestingly, Chiappe, Kenney and Smykowski (2003) found that the preference
of metaphors over preferred is largely due to the aptness, rather than the conventionality,
of the statement. They showed that what seems to predict the use of the metaphor form is
the extent to which the sentence is apt, i.e., it conveys important features of the topic.
The evidence in support of the career of metaphor account suggests that some
novel tropes (perhaps those that are not particularly apt, as the current investigation will
argue) may initially be processed as comparisons/similes and that only once they become
lexicalized are they understood as categorizations. However, on the basis of the studies
reported by Bowdle and Gentner (2005), one cannot conclude that all novel metaphors
begin their “career” as similes. Indeed, anecdotal evidence suggests that some novel
tropes are introduced and only make sense as metaphors, e.g., “What will be the next
Enron?”, not as similes, e.g., “What will be like the next Enron?”
Moreover, Glucksberg and Haught (2006) found that some novel metaphors can
be preferred and understood more quickly as categorizations than as comparisons. In this
experiment, the novel metaphors generated by Bowdle and Gentner (2005), e.g., “A
rumor is a mushroom”, “Moonlight is bleach” were our literal-referent metaphors. We
modified their vehicles by an adjective that was only applicable to the metaphor topic,
which yielded our figurative-referent metaphors, e.g., “A rumor is an unfounded
mushroom” and “Moonlight is romantic bleach”. Aptness ratings and comprehension
times revealed a preference for the comparison over the categorization form only for
DIFFERENCES BETWEEN METAPHORS AND SIMILES
15
15
literal-referent metaphors. For figurative-referent metaphors, the reverse was true: they
were judged as more apt and understood more quickly in categorical than in comparison
form.
An Overview of the Present Research
A critical assumption of the comparison and career of metaphor accounts is that
similes and metaphors convey the same meaning and are interchangeable. This article
challenges this assumption. I argue that metaphors and similes are processed and
interpreted differently and they are not always interchangeable.
The first proposal is that metaphors and their corresponding similes could differ in
their interpretations. This hypothesis is contrary to the comparison and career of
metaphor accounts, which assume equivalence in meaning of the two tropes. Experiment
1 was designed to test a likely preference for metaphorical properties in the
interpretations of metaphors and a preference for literal vehicle properties in the
interpretations of similes.
The second prediction is that some novel tropes could be privileged in metaphor
over simile form. This stands in contrast to the career of metaphor account, according to
which all novel metaphors are processed as similes. To test this hypothesis, apt novel
tropes were generated, a method that allowed for the study of differences between novel
and conventionalized metaphors and similes without concern for the confounding
variable of aptness. Data from aptness ratings (Experiment 2), comprehensibility ratings
(Experiment 3) and response times in a sensibility judgment task (Experiment 4) speak
directly to this hypothesis.
DIFFERENCES BETWEEN METAPHORS AND SIMILES
16
16
Experiment 1
In general, metaphors and similes have similar interpretations, so they can be paraphrased
into each other. However, via dual-reference, metaphors are understood to refer to a
higher level of abstraction than similes do. Consider the lawyers-sharks metaphor and its
corresponding simile. Whereas the simile predicate sharks refers only to the literal level
of the term, with its characteristic properties of viciousness and aggressiveness, the
metaphor vehicle denotes an abstract, superordinate category, that can include members
other than the prototypical marine creature shark. Therefore, in the metaphor form, the
topic lawyers is included into the abstract category that represents all vicious and
predatory things and people. This metaphoric category elicits additional properties that
emerge from the interaction of topic and vehicle and that could be ascribed to vicious
people and to the topic term lawyers, but not to the literal shark. The role of such
emergent features has been documented in the context of the poetic appreciation of
metaphors (Utsumi, 2005).
One testable implication of this dual-reference account is that metaphors should
elicit interpretations that are richer than literal interpretations of similes. For the lawyer-
shark example, emergent properties such as competent, which are not typical of the literal
vehicle shark, but are elicited in the context of the metaphor, should be preferred in
describing the meaning of the metaphor than the simile.
Barsalou (1982) proposed that certain properties of concepts, termed context-
independent, are, as the name suggests, always activated, regardless of contextual
relevance, whereas other properties are context-dependent, i.e., they are salient only in a
given context. For instance, the property round for the concept basketball is activated
DIFFERENCES BETWEEN METAPHORS AND SIMILES
17
17
regardless of context. A property such as it floats becomes activated only in certain
contexts, e.g., in the assertion The basketball fell into the pool. Becker (1997) further
categorized the properties that may or may not be activated during metaphor
comprehension into four classes: features that are common to both topic and vehicle;
vehicle-shared features; topic-shared features; and so-called emergent features, a
concept similar to Barsalou’s context-dependent properties. In her study, Becker found
that when asked to interpret metaphors the participants listed more vehicle-shared and
emergent features than either common or topic-shared features. Moreover, emergent
features seemed to be most influenced by the vehicle representation, as shown in an
analysis of the context-dependent features: more similar features were listed for sets of
metaphors with the same vehicle than for sets of metaphors that shared the same topic.
This study replicated the results found by Tourangeau and Rips (1991) with different
items and a modified experimental design.
Further support for the activation of these emergent features comes from a study
by Gineste, Indurkhya and Scart (2000), which showed that over 60% of the properties
generated during the interpretation of a metaphor were context-dependent, i.e., not
directly associated with either the topic or the vehicle. As Glucksberg, Manfredi and
McGlone (1997) pointed out, people seem to arrive at an appropriate interpretation of a
metaphor via an interaction between the topic and the vehicle.
Unlike the comparison and career of metaphor models, the categorization account
postulates that metaphors and similes may yield interpretations that are different. It
allows for some overlap in interpretations because the metaphor vehicle refers to both the
basic, literal category and the superordinate, abstract one. The account predicts that
DIFFERENCES BETWEEN METAPHORS AND SIMILES
18
18
interpretations of metaphors can include different properties that emerge primarily in the
context of the metaphor, hence they may not be derived from the corresponding simile.
Specifically, interpretations of metaphors such as “Some ideas are diamonds”,
whose vehicles do not refer directly to a literal entity, but to a superordinate concept (i.e.,
valuable and desirable things), should include more metaphorical or emergent properties,
such as creative and insightful. These properties are normally not included in the
representation of the literal concept diamonds. They are activated only in the context of
the metaphor. Interpretations of similes, e.g., “Some ideas are like diamonds”, on the
other hand, should include fewer of these emergent properties and more vehicle
properties, such as valuable and rare, that do apply to diamonds regardless of context.
Method
Participants This study consisted of two parts. In the first part, seven native speakers of
English rated adjectives on their applicability to the metaphor vehicles. Twenty-six
Princeton University undergraduates participated in the experiment proper for course
credit.
Materials Twenty-four metaphors were selected that could also be expressed as similes.
For each metaphor, the author generated and, following a norming study, selected three
adjectives describing vehicle properties and three adjectives describing emergent
properties. Each adjective was then incorporated in a sentence that described an
interpretation of the metaphor. For example, for “Some lawyers are sharks”, the
DIFFERENCES BETWEEN METAPHORS AND SIMILES
19
19
assertion “Some lawyers are vicious” is an interpretation that uses a vehicle property.
“Some lawyers are greedy” is an interpretation that uses an emergent property.
Norming: The selection of the properties was based on ratings from seven judges, who
rated on a scale from 1 to 7 how well each of the nine adjective generated by the
experimenter (three vehicle-applicable, three emergent and three fillers) applied to the
target concept, i.e., the metaphor vehicle. As expected, vehicle adjectives were rated as
significantly better descriptors of the vehicle concept (M = 5.05, SD = .42) than emergent
adjectives (M = 3.14, SD = .62) or fillers (M = 2.75, SD = .69), p < .0001.
Design and Procedure A 2 x 2 within-subjects design was used: statement format
(metaphor or simile) by type of property (vehicle or emergent). The order of the items
was randomized. Participants were asked to rate on a scale from 0 (“not at all”) to 6
(“very well”) how well they thought a particular interpretation fit with their
understanding of a given metaphor. It was emphasized that the ratings should reflect the
participant’s own judgment and that there were no right or wrong answers. Four practice
items followed the instructions to familiarize the subjects with the task and with the
presentation format.
Results and Discussion
Two repeated measures ANOVAs were conducted: one with participants as a random
factor (Fp) and one with items as a random factor (Fi). The participants ANOVA
revealed that vehicle properties (M = 3.88, SD = .66) received higher ratings than
DIFFERENCES BETWEEN METAPHORS AND SIMILES
20
20
emergent properties (M = 3.36, SD = .65), both for metaphors, Fp (1, 25) = 19.9, p < .001
and for similes, Fp (1, 25) = 38.2, p < .0001. The items ANOVA showed that the effect
of property type was significant for similes, Fi (1, 23) = 11.49, p < .003, but not for
metaphors, Fi (1, 23) = 2.67, p > .11.
These findings are consistent with the dual reference account: the metaphor
vehicle and the simile predicate share the same literal referent, e.g., the fish shark in the
example “His lawyer is (like) a shark”. Therefore, interpretations of both the metaphor
and the simile should include properties that describe the literal referent of the metaphor
vehicle. More importantly, the dual reference mechanism predicts that emergent
properties should be preferred in the interpretations of metaphors than in the
interpretations of similes. This is exactly what we found: emergent properties were rated
significantly higher for metaphors (M = 3.48, SD = .62) than for similes (M = 3.25,
SD=.71), Fp(1, 25)=14.7, p<.001 and Fi (1, 23)=5.22, p<.03.
Figure 1: Ratings of interpretations with vehicle and emergent properties as a function of
statement form: metaphor and simile, Experiment 1.
3
3.2
3.4
3.6
3.8
4
4.2
Emergent Vehicle
Rat
ings
Property Type
Metaphor Simile
DIFFERENCES BETWEEN METAPHORS AND SIMILES
21
21
Figure 1 shows the crucial significant interaction, Fp (2, 50) = 12.7, p <. 001 and
Fi (2, 46) = 8.38, p < .008, between statement form (simile vs. metaphor) and type of
property (vehicle vs. emergent): the interpretations that included vehicle adjectives were
rated as equally applicable to similes (M = 3.93, SD = .67) and to metaphors (M = 3.83,
SD = .64), whereas the interpretations that included emergent adjectives received higher
ratings for metaphors (M = 3.48, SD = .62) than for similes (M = 3.24, SD = .71).
These results speak against the original comparison account, which holds that
metaphors are implicit similes, and they support the predictions made by the
categorization model and its postulated dual reference mechanism.
The preference of interpretations with emergent attributes for metaphors than for
similes supports the predictions made by the categorization model and its postulated dual
reference mechanism. Whereas the simile predicate sharks refers only to the literal level
of the term, with its characteristic properties of viciousness and aggressiveness, the
metaphor vehicle denotes an abstract, superordinate category, that can include members
other than the prototypical marine creature shark. Therefore, for the metaphor “Some
lawyers are sharks”, the topic lawyers is included into the abstract category that
represents all vicious and predatory things and people. This metaphoric category elicits
additional properties that emerge from the interaction of topic and vehicle and that could
be ascribed to vicious people and to the topic term lawyers, but not to the literal shark.
The findings clearly speak against the original comparison account, which holds
that metaphors are implicit similes. They support the categorization model, but they
could also be accommodated by the career of metaphor argument, which postulates that
DIFFERENCES BETWEEN METAPHORS AND SIMILES
22
22
metaphors acquire dual reference and are treated as categorizations as they become
lexicalized or commonly used in everyday language. As metaphors become
conventionalized, they could convey meanings that are richer and more complex than
those conveyed by similes (during the initial comparison process). Therefore,
conventional metaphors, which can be processed as categorizations, may be somewhat
privileged in interpretations over novel metaphors, which are processed as
comparisons/similes. Experiments 2 through 4 aimed to draw a clear distinction between
the categorization and the career of metaphor models by further exploring potential
differences between metaphors and similes.
Experiment 2
The dual reference mechanism that explains the findings in Experiment 1 is at the
core of the categorization model (Glucksberg & Keysar, 1990), but it also plays a critical
role in the career of metaphor model (Bowdle & Gentner, 2005). The latter account posits
that novel metaphors, i.e., those that have just been introduced in a language, are always
understood as comparisons. Then, through repeated and consistent use, the metaphors
become conventionalized and their vehicles acquire polysemy or dual reference.
Bowdle and Gentner supported their account by the finding that novel metaphors
were strongly preferred in simile form, whereas conventional assertions were slightly
preferred in metaphor form. One important variable that may have been confounded with
the assessment of novelty/conventionality, however, is the aptness of the expressions: the
novel metaphors used, e.g., “A fisherman is a spider”, “A mind is like a kitchen”,
DIFFERENCES BETWEEN METAPHORS AND SIMILES
23
23
“Science is a glacier”, were rated as significantly less apt than their conventionalized
counterparts.
To address the challenge of generating apt novel metaphors, existing
conventionalized metaphors such as “His lawyer is a shark” were modified by an
adjective, such that the new expressions had vehicles formed not by one noun (shark), but
by a conceptual combination consisting of an adjective and a noun (e.g., old shark,
respectable shark, razor-toothed shark). Using this procedure, experiments 2, 3, and 4
explored how metaphors and similes are interpreted when they include one of three types
of adjectives. The adjectives differed as a function of whether the properties they
conveyed applied in their literal sense only to the topic term, only to the vehicle term, or
to both the topic and the vehicle.
The prediction was that novel tropes with adjectives applicable to the topic only,
e.g., “Some lawyers are well-paid sharks”, are privileged in metaphor form. When
converted into similes, e.g., “Some lawyers are like well-paid sharks”, they become less
apt. If the results support this hypothesis, then they would provide a clear counterexample
to the career of metaphor argument. They would show that novel tropes can be
privileged in categorical (metaphor) rather than comparison (simile) form. Hence, novel
metaphors are not necessarily preferred as comparison assertions. The findings would
support the categorization model, whose dual reference mechanism allows for processing
and interpretation differences in metaphors versus similes. The metaphor Some lawyers
are well-paid sharks is apt because the vehicle refers to the metaphorical shark, which
can be described as well-paid. In the corresponding simile, sharks is used at its literal
DIFFERENCES BETWEEN METAPHORS AND SIMILES
24
24
level, in which case the attribute well-paid used to describe the fish would make for a
strange conceptual combination at best.
For the other types of adjectives, the predictions regarding differences in aptness
between the metaphor and the simile conditions are as follows. Statements with topic-
and-vehicle-applicable adjectives (e.g., “Some lawyers are old sharks”) should be
equally apt in metaphor and simile form. Metaphors with vehicle-applicable adjectives
(e.g., “Some lawyers are razor-toothed sharks”) should be judged as apt as the
corresponding similes, but may be judged as less apt than the unmodified source
metaphors (e.g., “Some lawyers are sharks”) because the adjective does not apply to the
metaphor topic.
Method
Participants The experiment consisted of two parts. Sixteen Princeton undergraduates
were given course credit for participating in the norming part of the experiment, which
determined the three adjectives to modify each metaphor’s vehicle term. Twenty-six
participants were then tested in the experiment proper for course credit. All the
participants were native speakers of English.
Materials The materials consisted of thirty-two pairs of nominal metaphors-similes.
Each metaphor vehicle/simile predicate was modified by an adjective or not. When an
adjective was included as a modifier, it was chosen from a norming study such that it was
applicable to the vehicle only, the topic only, or both topic and vehicle.
DIFFERENCES BETWEEN METAPHORS AND SIMILES
25
25
The experimenter selected the three target attributes for each metaphor based on
the criteria specified above. For example, for the source metaphor Some lawyers are
sharks, which contains no adjective (NA), the topic-applicable (TA) adjective had to be
applicable to lawyers, but not to sharks (e.g., well-paid); the vehicle-applicable (VA)
adjective had to apply to sharks, but not to lawyers (e.g., razor-toothed), and topic-and-
vehicle-applicable (TVA) adjectives had to apply to both lawyers and sharks (e.g., old).
Then, in the norming part of the study, participants were asked to rate how well each of
six attributes applied to the representation of a concept. Three of these attributes were the
adjectives that had been generated by the experimenters and the other three were fillers.
The subjects rated the six adjectives on a scale from 0 to 6 based on how well each
property applied to the description of a noun. The nouns were the topic and the vehicle
terms, e.g., lawyer and shark. Half of the participants rated the applicability of the
properties to the topic and half of them rated the applicability to the vehicle. The nouns
and the adjectives were presented in a randomized order.
An initial 2 (target concept: topic and vehicle) x 3 (adjective type: TA, VA, TVA)
ANOVA revealed a significant interaction, Fi (2, 30) = 28.8, p < .0001 and Fp (2, 30) =
23.6, p < .0001. As predicted, topic-applicable (TA) adjectives were rated higher for the
topic (M = 5.04, SD = .71) than for the vehicle (M = 1.33, SD = .80). This effect was
reliable both by items, Fi (1, 31) = 287, p < .0001, and by participants, Fp = 83, p < .0001.
Vehicle-applicable (VA) adjectives were rated significantly higher for the vehicle (M =
4.97, SD = 1.08) than for the topic (M = 1.54, SD = 1.03), Fp (1, 31) = 71, p < .001 and Fi
(1, 31) = 282, p < .0001. This interaction of topic/vehicle and topic-applicable/vehicle-
applicable adjectives was significant, Fi (1, 31) = 565, p < .0001 and Fp (1, 31) = 380, p <
DIFFERENCES BETWEEN METAPHORS AND SIMILES
26
26
.0001. There was no difference in ratings for the TVA adjectives, which were equally
applicable to the topic (M = 4.4, SD = 1.15) and the vehicle (M = 4.75, SD = 1.07), p >
.3.
Design and Procedure A 2 x 4 within subjects design was used: statement format (simile
or metaphor) by type of adjective (topic-applicable, vehicle-applicable, topic-and-vehicle
applicable, or no adjective). Half of the statements were presented as metaphors in one
block and the other half were presented as similes in a different block. The order in which
the two blocks were presented was counterbalanced and order of statements within each
block was randomized. The task consisted of an aptness measure, in which the
participants rated how good each metaphor or simile was on a scale from 1 (not at all apt)
to 10 (extremely apt).
Results and Discussion
The experimental design aimed to avoid the confounding issue of aptness that was
problematic in the studies reported by Bowdle and Gentner (2005). Instead of using
novel figuratives that are less apt than the conventional ones, we employed novel,
adjective-modified metaphors that were judged to be as apt as their conventional,
unmodified counterparts. Indeed, there was no difference between the aptness ratings of
novel topic-applicable metaphors (M = 6.2, SD = 1.51) and their non-modified
conventional counterparts (M = 6.7, SD = 1.41), p > .23. In simile form, the topic-
applicable statements were rated as significantly less apt (M = 5.25, SD = 1.65) than the
corresponding non-modified similes (M = 6.5, SD = 1.56), Fp (1, 31) = 22.06, p < .0001
DIFFERENCES BETWEEN METAPHORS AND SIMILES
27
27
and Fi (1, 31) = 17.8, p < .001 and in fact lower than the ratings of any of the other
conditions.
As predicted, the similes modified by topic applicable adjectives were rated as
less apt (M = 5.25, SD = 1.65) than the corresponding metaphors (M = 6.2, SD = 1.51).
This effect was reliable by participants, Fp (1, 31) = 9.21, p < .005, and by items, Fi (1,31)
= 6.51, p < .016. For the unmodified condition, there was no difference between
metaphors (M = 6.7, SD = 1.41) and similes (M = 6.5, SD = 1.56), p > .97. The 2
(metaphor / simile) x 2 (topic-applicable / non-modified) interaction was significant in
the ANOVA by items, Fi (2, 30) = 4.3, p < .05 and marginally significant in the analysis
by participants, Fp (2, 30) = 3.6, p = .06. This finding undermines the career of
metaphor’s assumption that all novel tropes are understood as comparisons, hence they
should always be preferred in simile form. What seems to determine this preference is
not level of conventionalization, but aptness. The novel simile “Some lawyers are like
well-paid sharks” is not very apt, but the novel metaphor “Some lawyers are well-paid
sharks” is. This result supports the dual reference hypothesis: a metaphorical shark, to
which the superordinate metaphor category refers, can be well-paid (a TA adjective), but
the literal shark, to which the simile predicate refers, cannot.
In the other two conditions (VA and TVA), the novel metaphors were not rated as
significantly different in aptness from their simile counterparts. The metaphors modified
by a topic-applicable adjective and those modified by a topic-and-vehicle-applicable
adjective were also rated as apt as the source (NA) metaphors.
There was a difference between the VA metaphors (M = 5.74, SD = 1.47) and the
NA metaphors (M = 6.70, SD = 1.41): the metaphors with vehicle-applicable adjectives
DIFFERENCES BETWEEN METAPHORS AND SIMILES
28
28
were rated significantly lower than the non-modified metaphors, Fp (1, 31) = 13.42, p <
.001 and Fi (1,31) = 6.3, p < .017. This finding may explained by the fact that a vehicle-
applicable adjective, such as razor-toothed for the lawyer/shark example, reflects an
essential property of the metaphor vehicle. Therefore, it may have been perceived as
redundant when used to modify the vehicle: because the property razor-toothed is
intrinsic to a description of any shark, people preferred the more parsimonious of the two
categories: sharks over razor-toothed sharks.
This experiment provided direct evidence against the career of metaphor
account’s claim that all novel metaphors are privileged in simile form. Bowdle and
Gentner (2005) attribute the simile preference for the novel tropes to the fact that all new
metaphors are processed as comparisons and they dismiss the potential role that aptness
may play. The current experiment presented a counterexample where a novel trope that
was as apt as a conventional one was preferred in metaphor over simile form. As
expected, the novel topic-applicable metaphors were rated as apt as their non-modified
conventional counterparts, but the corresponding topic-applicable similes were rated as
less apt than the non-modified similes and indeed lower than any of the other categories
of statements.
Experiment 3
In Experiment 2, the results showed that some novel metaphors (whose vehicles
were modified by topic-applicable adjectives) were considered more apt when presented
in categorical than in simile form. The aptness test was important because it showed that,
in contrast to the career of metaphor account, not all novel metaphoric expressions are
DIFFERENCES BETWEEN METAPHORS AND SIMILES
29
29
privileged in simile than in metaphor form. Moreover, the comparison of the novel
(adjective-modified) metaphors with the conventionalized (unmodified) ones revealed
that the two groups of statements did not differ reliably in their aptness. This finding
ensured that any preference for any of the novel items in metaphor or in simile form was
not due to differences in their aptness, a concern that had been raised and not adequately
addressed by Bowdle and Gentner (2005).
Experiment 3 aimed to extend the findings from Experiment 2, using a task that
required participants to judge the comprehensibility, rather than the aptness of a set of
metaphors. This test is an important complementary tool to the aptness ratings because it
allows for an assessment of the ease (or difficulty) with which the participants understand
the items. By employing the same materials that were used in Experiment 2, the study
can also address the possible criticism that some of the novel items, though apt and
preferred in metaphor form, might have been understood with more difficulty than their
corresponding similes.
The predictions for the current experiment mirror those of Experiment 2. Some
novel apt tropes can be preferred in metaphor than in simile form. An interaction
between type of trope and vehicle modifier was predicted: the similes whose predicates
are modified by topic-applicable adjectives should be rated significantly lower on the
comprehensibility scale than their corresponding metaphors, but there should be no
difference in ratings for the unmodified tropes.
Method
Participants Thirty-two Princeton undergraduates, all native English speakers,
participated in this study for course credit.
DIFFERENCES BETWEEN METAPHORS AND SIMILES
30
30
Materials The items were identical to those used in Experiment 2.
Design and Procedure The design and the procedure were parallel to those in
Experiment 2. A 2 x 4 within subjects design was used: statement format (simile or
metaphor) by type of adjective (topic applicable, vehicle applicable, topic and vehicle
applicable, or no adjective). The task consisted of rating each statement based on how
easy it is to understand it on a scale from 1 (not at all easy) to 10 (extremely easy). The
participants were asked to use their best judgment and to produce the first rating that
came to mind, without going back and revising their responses.
Results and Discussion
Analyses of variance were conducted with both participants and items as random factors.
An initial 2 (statement form: simile or metaphor) X 4 (property type: TA, VA, TVA or no
adjective) ANOVA revealed a significant interaction, Fp (1, 29) = 4.7, p < .009 and Fi (1,
29) = 5.3, p < .001.
As predicted, the metaphors that included topic-applicable adjectives received
higher ratings of comprehensibility (M = 6.66, SD = 1.61) than the corresponding similes
(M = 5.72, SD = 1.78), Fp (1, 31) = 7.59, p < .01 and Fi (1, 31) = 11.92, p < .002. No
such difference was observed in the unmodified tropes, which were rated as equally
comprehensible regardless of whether they were presented as metaphors (M = 7.27, SD =
1.52) or as similes (M = 7.35, SD = 1.59). This crucial interaction was significant both
by participants Fp (1, 31) = 5.44, p < .026 and by items, Fi (1, 31) = 5.19, p < .03. The
ratings for topic-applicable similes were lower (M = 5.72, SD = 1.78) than the ratings for
the corresponding non-modified similes (M = 7.27, SD = 1.52), Fp (1, 31) = 43.9, p <
DIFFERENCES BETWEEN METAPHORS AND SIMILES
31
31
.0001 and Fi (1, 31) = 36.2, p < .0001. However, the novel topic-applicable metaphors
were rated as not reliably different in comprehensibility from their non-modified
conventional counterparts.
The topic-applicable condition was the only one that yielded a significant
difference between metaphors and similes: there was no preference for either statement
form for any of the other metaphor types. As predicted, there was no reliable difference
between the novel metaphors that included topic-applicable or topic-and-vehicle-
applicable adjectives and the original conventional metaphors: both types of statements
were rated as being equally comprehensible. This point speaks against the career-of-
metaphor hypothesis, which predicts that novel figuratives should always be more
comprehensible in simile than in metaphor form.
Experiment 4
In order to converge on the argument that novel metaphors can be privileged in
categorical form, Experiment 4 employed an online measure of comprehension using the
same set of materials as in Experiments 2 and 3. The task did not require the participants
to produce or make decisions about interpretations. Instead, they were simply asked to
read various statements (the target metaphors and literal fillers) and make a speeded
judgment about each assertion’s comprehensibility, i.e., whether it makes sense or not.
Overall, the participants should take longer to understand the novel tropes than the
conventionalized ones because the novel statements include an additional word (the
adjective), which increases reading time, and because they should be less familiar than
the conventional items. Participants should also take more time to reach a decision
DIFFERENCES BETWEEN METAPHORS AND SIMILES
32
32
regarding the similes with topic-applicable adjectives and they should reject them more
often, i.e., produce more “no” responses, because these assertions are less apt, as was
shown in Experiment 2. No significant differences between metaphors and similes are
expected for any of the other conditions.
The prediction is straightforward: similes with topic-applicable adjectives, e.g.,
“Some lawyers are like well-paid sharks”, would not only be judged to be less apt and
comprehensible, as shown in Experiments 2 and 3, but they would also take longer to
understand than their corresponding metaphors, e.g., “Some lawyers are well-paid
sharks”.
Method
Participants Twenty-four Princeton undergraduates, all native speakers of
English, participated for course credit.
Materials The materials consisted of the same set of thirty-two nominal
metaphor-simile pairs that were used in Experiments 2 and 3. In addition, sixty-four
literal statements were included as fillers: half of them were comprehensible and the
other half were not. Sixteen of the literal comprehensible items and sixteen of the
incomprehensible items were of the form X is a AY, where X and Y are nouns and Y is an
adjective, and the rest of the fillers had other grammatical structures, such as X [transitive
verb] AY (e.g., X had a AY). Twelve additional items were generated and were used as
warm up materials at the beginning of the experiment.
DIFFERENCES BETWEEN METAPHORS AND SIMILES
33
33
Design and Procedure This experiment employed a sentence comprehension task, in
which I manipulated the statement format and the type of modifier. This yielded, as in
the previous two studies, a 2 x 4 design: simile or metaphor by vehicle-applicable (VA)
adjective, topic-applicable (TA) adjective, topic-and-vehicle-applicable adjective (TVA)
or no adjective (NA).
The statements were presented in random order on a computer screen and the
participants were asked to complete a sentence comprehension task. They were
instructed to look at a fixation cross, which always appeared in the center left area of the
screen. After 1500 milliseconds, a statement appeared, which the participants were asked
to read as soon as they saw it on the computer screen. They were instructed to make a
judgment as quickly and accurately as possible about whether the statement made sense
or not. They responded by pressing one of two keys, each of which was clearly labeled
with yes or no. The instructions emphasized that the participants keep constant steady
hand positions, which was assessed during the practice trials, in order to minimize the
variability attributed to different hand movements.
Ten practice trials were used to familiarize the participants with the task, and the
first twelve items in the experiment proper were considered warm-up. All the statements
were presented in random order, and the computer recorded the participants’ response
choices (yes or no) and their reaction times.
Results
Participant (Fp) and item (Fi) analyses of variance were conducted on the time it took
each participant to decide on the comprehensibility of each statement and on the
percentage of yes and no responses that they produced.
DIFFERENCES BETWEEN METAPHORS AND SIMILES
34
34
Response times As predicted and as shown in Figure 2, the mean response times were
longer for the novel statements that included an adjective (M = 1995.84 ms., SD =
627.77) than for the unmodified, conventional ones (M = 1604.75 ms., SD = 605.31).
This result is attributable to two factors: first, the novel tropes included an additional
word, which may have increased reading times. Second, the modified tropes were novel
and less familiar than the conventionalized tropes.
Figure 2: Response times in milliseconds as a function of statement form: metaphor and
simile, and of property/adjective type: no adjective (NA), topic-applicable (TA), vehicle-
applicable (VA), and topic-and-vehicle-applicable (TVA), Experiment 4.
The participants took significantly longer to respond to similes with topic-applicable
adjectives (M = 2243.45, SD = 143.36) than to their metaphor counterparts (M =
1000
1200
1400
1600
1800
2000
2200
2400
2600
NA TA VA TVA
Res
pons
e Ti
mes
(ms.)
Property Type
Metaphors Similes
DIFFERENCES BETWEEN METAPHORS AND SIMILES
35
35
1946.32, SD = 121.59), Fp (1, 23) = 9.49, p < .005 and Fi (1, 31) = 8.3, p < .008. No such
difference between similes and metaphors was observed for any of the other adjective
conditions. Three separate 2 (metaphor or simile) X 2 (topic-applicable adjective or one
of the other adjectives) ANOVAs were conducted and they revealed the following
significant interactions. While the similes with TA adjectives took longer to understand
than the TA metaphors, VA similes and metaphors were understood as quickly, Fp (1, 23)
= 5.6, p < .027, Fi(1, 31)=4.9, p<.03. Similarly, there was no difference in response times
to TVA metaphors and similes, Fp(1, 23)=4.36, p<.048 and Fi(1, 31) = 5.23, p < .04.
These findings are consistent with the aptness and comprehensibility results from
Experiments 2 and 3, respectively. When a simile predicate is modified by a topic-
applicable adjective, participants take longer to understand the simile than they do to
understand the corresponding metaphor. The reason for the delay is that similes such as
“Some lawyers are like well-paid sharks” are not particularly apt. The adjective well-
paid does not apply to the literal term sharks, to which the simile predicate refers. On the
other hand, well-paid does apply to the metaphor topic lawyers and to the metaphoric
vehicle referent sharks. This dual reference accounts for the difference in aptness and
comprehensibility and for the delay in understanding the simile over the metaphor.
Sensibility judgments Overall, the participants responded that the similes made sense
more often than for metaphors, with the crucial exception of the case when a topic-
applicable adjective was included. Although the difference for the topic-applicable
condition between metaphors (M = .76, SD = .21) and similes (M = .72, SD = .25) was
not significant, p > .43, there was a significant interaction with the vehicle-applicable
DIFFERENCES BETWEEN METAPHORS AND SIMILES
36
36
condition, in which similes (M = .77, SD = .22) were judged to make sense more often
than their corresponding metaphors (M = .65, SD = .29), Fp (1, 23) = 4.28, p < .05. This
interaction was marginally significant by items, Fi (1, 31) = 2.8, p > .07. The interaction
of the topic-applicable adjective with the no adjective condition was marginally
significant by participants, Fp (1, 23) = 3.76, p = .065 and by items, Fi (1, 31) = 3.21, p <
.08.
These results were consistent with our predictions and with the ratings of aptness
and comprehensibility from Experiments 2 and 3. First, similes whose vehicles were
modified by topic-applicable adjectives took longer to understand than their
corresponding metaphors. Second, while the similes were judged to make sense more
often than the metaphors for the vehicle-applicable, topic-and-vehicle-applicable, and no
adjective conditions, the reverse tendency was observed for the topic-applicable
condition.
General Discussion
Four experiments made it possible to distinguish between existing models of metaphor
comprehension and to draw some important observations about the processing of
adjective-noun combinations in a figurative context.
First, they showed that novel tropes could be privileged in metaphor form, a
finding that the career of metaphor argument cannot account for. When novel metaphors
such as “Some lawyers are well-paid sharks” include adjectives that apply only to the
metaphor topic, they are considered more apt and comprehensible and take less time to
process than their corresponding similes, e.g., “Some lawyers are like well-paid sharks”.
DIFFERENCES BETWEEN METAPHORS AND SIMILES
37
37
The dual reference function of the vehicle term provides a straightforward account for
this finding: while the metaphorical shark can be well-paid, the literal one, to which the
simile refers, cannot. Such results support the categorization model by showing that
novel metaphors can be privileged as categorical assertions. The findings are a clear
counterexample to the career of metaphor’s postulate of a mandatory comparison process
for all novel metaphors.
Second, the current findings showed that metaphors whose vehicles are modified
by certain adjectives could express meanings that differ from the meanings of the
corresponding similes. For instance, the interpretation “Some lawyers are vicious” was
rated to be equally applicable to the metaphor and the simile form of “Some lawyers are
(like) sharks”, whereas “Some lawyers are greedy” was preferred as an interpretation of
metaphors. So, when the interpretation included properties such as greedy, which do not
apply to the literal shark, but emerge in the context of the metaphor, it was the preferred
to describe the meaning of the metaphor, but not the simile form. Such differences in
interpretations can be accounted by the dual reference property of the vehicle: the
adjective vicious refers to the literal predicate, encountered in the simile, while greedy
refers to the abstract, metaphorical category, encountered in the metaphor. Interestingly,
recent fMRI data suggest that different brain areas may be activated during the
processing of metaphors and similes (Shibata, Toyomura, Motoyama, Itoh, Kawabata, &
Abe, 2012).
Third, the use of an adjective modification paradigm yielded conceptual
combinations that served as novel metaphor vehicles, e.g., old sharks in the lawyers-
sharks example. This procedure raises the interesting and previously unexplored question
DIFFERENCES BETWEEN METAPHORS AND SIMILES
38
38
of how such adjective-noun combinations are processed in literal versus figurative
expressions. The issue has been marginalized by theories of conceptual combinations,
which are focused largely on nouns and modifiers that are not embedded in sentences.
The results reported here suggest that some adjective-noun combinations may be
understood differently or may elicit different meanings depending on the type of
sentential or discourse context in which they are presented. Specific interpretations seem
to be computed online and require an enriched form of semantic composition (e.g.,
Traxler, Martin, & McElree, 2002). Therefore, the same conceptual category may be
interpreted differently in a literal sentence or in a simile, than in the corresponding
metaphor.
Theoretical Implications for Metaphor Comprehension
Existing models of metaphor comprehension, including the standard comparison
models (e.g., Ortony, 1979) and the recent career of metaphor argument (Bowdle &
Gentner, 2005), assume that metaphors and similes are interchangeable. The dual
reference property of the vehicle term is key. It specifies that metaphors and similes have
distinct referents, which can lead to distinct meanings and distinct comprehension
processes: categorization for metaphors and comparison for similes.
The career of metaphor account acknowledges the role of dual reference: as novel
metaphors are lexicalized, their vehicle terms become polysemous, i.e., they are used to
refer to both the literal term and the abstract, metaphorical category. However, crucially,
dual reference is not an a priori property of the vehicle. Instead, it is acquired after
repeated use as a byproduct of the initial comparison process and is only relevant for
DIFFERENCES BETWEEN METAPHORS AND SIMILES
39
39
conventional metaphors. Therefore, this account cannot accommodate the current results,
which show significant changes in the meaning of novel metaphors as a function of
grammatical form. The postulated shift from comparison (similes) to categorization
(metaphors) that occurs over time implies equivalence in the meaning of the two tropes.
The career of metaphor argument also cannot explain the aptness and
comprehensibility effects reported here. Bowdle and Gentner (2005) argue that novel
metaphors are invariably processed as comparisons, hence they should always be
privileged in simile form. Conventional metaphors are processed as categorizations or
comparisons, hence they may be privileged in metaphor form. This is not always the
case. We showed that novel metaphors could be privileged when expressed in metaphor
rather than simile form, as shown in the current experiments. Conventional metaphors
also exist that are apt and interpretable as similes, e.g., “The lecture hall was so crowded
that she felt like a sardine,” but not as metaphors, e.g., She felt she was a sardine.
According to Bowdle and Gentner (2005), conventionality is a sufficient
condition for categorization. However, based on the evidence reported here, it is not a
necessary one. Instead, the present findings support the view that apt metaphors, both
novel and conventionalized, can be understood as categorical, class-inclusion statements,
while metaphors that are not apt, as well as similes, are likely to require a comparison
evaluation. The key factor that determines whether metaphors are processed as
comparisons or categorizations seems to be not conventionality, but aptness. Findings by
Chiappe et al. (2003) similarly suggest that aptness, rather than conventionality, mediate
the categorization process via which metaphors are understood.
DIFFERENCES BETWEEN METAPHORS AND SIMILES
40
40
Aptness is viewed as relative difficulty of comprehension, such that apt metaphors
are easily understood, even out of context, whereas non- or less apt metaphors are more
difficult to process, and may require non-trivial inferential work. In other words,
metaphors differ in comprehensibility, with aptness being one determinant, and
familiarity another, along with possibly other variables that may affect comprehensibility.
The issues of aptness and conventionality are related, in that aptness can be an
important predictor of a metaphor’s “career” and a prerequisite for “promotion” to
metaphor conventionalization. Bowdle and Gentner (2005) define conventionality in
terms of metaphors and similes with vehicle terms “that have become polysemous
because of repeated and consistent figurative use” (pp. 204). Aptness is not directly
incorporated in this definition, even though in most cases, aptness seems to be a
necessary, albeit not sufficient, condition for conventionalization. Metaphoric
expressions such as “Mind is a kitchen” and “Science is a glacier”, which are not
particularly good or apt to begin with, are unlikely to become lexicalized. Mere exposure
to the same vehicle term, which was used to simulate lexicalization in one of Bowdle and
Gentner’s experiments, does not seem to ensure or predict conventionalization.
Nevertheless, it is conceivable that some metaphors that are not very apt when they enter
a language could persevere: with repeated use, they can climb up the career ladder to
become more apt and acceptable as categorization statements. But, most importantly
from a theoretical standpoint, novel metaphors do not have to start out as comparisons
and follow the lexicalization route to be promoted to categorization. Instead, as the
current studies have shown, they can be introduced and interpreted as categorizations
right from the beginning of their career.
DIFFERENCES BETWEEN METAPHORS AND SIMILES
41
41
The implications of the present findings are clear. First, aptness seems to be a
more important condition than conventionality for metaphors to be understood as
categorical assertions. Apt metaphors, regardless of whether they are novel and
conventionalized, are processed as categorizations, whereas similes and inapt metaphors
are likely to be processed as comparisons. Second, metaphors and similes are not
interchangeable and they do not always mean the same thing. Therefore, theories that
assume equivalence in the interpretations of metaphors and their corresponding similes
are flawed.
DIFFERENCES BETWEEN METAPHORS AND SIMILES
42
42
References
Bowdle, B. F., & Gentner, D. (2005). The career of metaphor. Psychological Review,
112, 193-216.
Chiappe, D., Kennedy, J. M., & Smykowski, T. (2003). Reversibility, aptness, and the
conventionality of metaphors and similes. Metaphor and Symbol, 18, 85-105.
Fogelin, R. J. (1988). Figuratively Speaking. New Haven: Yale University Press.
Gentner, D. (1983). Structure-mapping: A theoretical framework for analogy. Cognitive
Science, 7, 155-170.
Gentner, D., & Bowdle, B. F. (2001). Convention, form, and figurative language
processing. Metaphor and Symbol, 16, 223-247.
Gentner, D., & Clement, C. A. (1988). Evidence for relational selectivity in interpreting
analogy and metaphor. In G. H. Bower (Ed.), The psychology of learning and
motivation (pp. 307-358). New York: Academic Press.
Gentner, D., & Markman, A. B. (1997). Structure mapping in analogy and similarity.
American Psychologist, 52, 45-56.
Gibbs, R. W. (2011). Multiple constraints in theories of metaphor. Discourse Processes,
48, 575-584.
Gibbs, R. W., & Colston, H. L. (2012). Interpreting figurative meaning. New York, NY:
Cambridge University Press.
Gineste, M. D., Indurkhya, B., & Scart, V. (2000). Emergence of features in metaphor
comprehension. Metaphor and Symbol, 15, 117-135.
DIFFERENCES BETWEEN METAPHORS AND SIMILES
43
43
Glucksberg, S. (2001). Understanding figurative language: From metaphors to idioms.
London: Oxford University Press.
Glucksberg, S., & Haught, C. (2006a). On the relation between metaphor and simile:
When comparison fails. Mind and Language, 21, 360-378.
Glucksberg, S., & Haught, C. (2006b). Can Florida become like the next Florida? When
Metaphoric Comparisons Fail. Psychological Science, 17, 935-938.
Glucksberg, S., & Keysar, B. (1990). Understanding metaphorical comparisons: Beyond
similarity. Psychological Review, 97, 3-18.
Glucksberg, S., McGlone, M. S., & Manfredi, D. (1997). Property attribution in metaphor
comprehension. Journal of Memory and Language, 36, 50-67.
Grice, H. P. (1975). Logic and conversation. In P. Cole and J. Morgan (Eds.), Syntax
and Semantics, 3 (pp. 41-58). New York: Academic Press.
Haught, C. (in press). Spain is not Greece: How metaphors are understood. Journal of
Psycholinguistic Research.
Miller, G. A. (1979). Images and models: Similes and metaphors. In A. Ortony (Ed.),
Metaphor and Thought, (pp. 202-250). New York: Cambridge University Press.
Ortony, A. (1979). Beyond literal similarity. Psychological Review, 86, 161-180.
Searle, J. (1979). Metaphor. In A. Ortony (Ed.), Metaphor and Thought (pp. 92-123).
New York: Cambridge University Press.
Shen, Y. (1992). Metaphors and categories. Poetics Today, 13, 771-794.
Shibata, M., Toyomura, A., Motoyama, H., Itoh, H., Kawabata, Y., & Abe, J. (2012).
Does simile comprehension differ from metaphor comprehension? A functional
MRI study. Brain and Language, 121, 254-260.
DIFFERENCES BETWEEN METAPHORS AND SIMILES
44
44
Steen, G. J. (2007). Finding metaphor in grammar and usage: A methodological analysis of
theory and research. Amsterdam: John Benjamins.
Tourangeau, R., & Rips, L. (1991). Interpreting and evaluating metaphors. Journal of
Memory and Language, 30, 452-472.
Traxler, M. J., Martin, J. P., & McElree, B. (2002). Coercion in sentence processing:
Evidence from eye-movements and self-paced reading. Journal of Memory and
Language, 47, 530-547.
Tversky, A. (1977). Features of similarity. Psychological Review, 85, 327-352.
Utsumi, A. (2005). The role of feature emergence in metaphor appreciation. Metaphor
and Symbol, 20, 152–172.
Utsumi, A., & Sakamoto, M. (2011). Categorization as a process of predicative metaphor
comprehension. Metaphor and Symbol, 26, 299-313.